A new approach to detect
similar proteins from 2D gel
electrophoresis images
Nawaz Khan and Shahedur Rahman
Proceedings of the Third IEEE Symposium on
BioInformations and
BioEngineering(BIBE’03), 2003
Speaker: Ho, Yu An(何玉安)
Date: 2004/6/3
Outline
•
•
•
•
Introduction
Methodology
Experiments and results
Conclusion
nmlab
2
Introduction (1/4)
• GELLAB system (1981)
– Uses the point pattern comparison.
• MELANIE (1997)
– Compares spot clusters.
• Panek and Vohradsky (1999)
– Use the information from the neighbourhood
spots for comparison.
nmlab
3
Introduction (2/4)
• Although the spot in the source and target
image can be identical or similar, but still
the following parameters can very:
–
–
–
–
Background value
Protein spot intensity
Protein spot shape
Noise in the image
nmlab
4
Introduction (3/4)
Triosophosphate isomerase protein spots in two
nmlab
different images.
5
Introduction (4/4)
• This paper presents a novel approach for identifying the
identical or similar protein spot in 2D gel electrophoresis
images by considering the following factors:
– 2D gel electrophoresis protein spots differ significantly in two
different images even when they represent the same protein.
– Same or similar protein spots will lie at the same line of path
because of their electrophoresis mobility and molecular weight.
– The intensity of the matched regions in both images can be
different even thought it shows a correct matching.
– The region of similar spot at the target image must lie at the same
or different directional vector on the line of path.
nmlab
6
Methodology
1. Determining the position of the protein spot in
the source image
2. Defining the region of interest
3. Matching the selected protein spot in the target
image
4. Searching for the protein spot in the
neighbourhood area
5. Selecting the best matched spot
6. Retrieving 3D structure of a protein
nmlab
7
1. Determining the position of the
protein spot in the source image (1/3)
Source image divided into four quadrants.
nmlab
8
1. Determining the position of the
protein spot in the source image (2/3)
nmlab
9
1. Determining the position of the
protein spot in the source image (3/3)
Angle produced with the horizontal axis for
any point of interest on the vertical plane.
nmlab
10
2. Defining the region of interest
(1/2)
nmlab
11
2. Defining the region of interest
(2/2)
(a) A set of points defined by the user, (b)
Defining the region of interest.
nmlab
12
3. Matching the selected protein spot
in the target image
The target point will be considered as ' matched point' when :
1.  s  T ;  s   T ; M s  M T
In other situation where
2.  s  T ;  s   T ; M T  M s or M T  M s (where M T is within
a threshold value), it will be considered as ' spot found' and it
might be the same or similar protein.
nmlab
13
4. Searching for the protein spot in
the neighbourhood area (1/2)
Non emptied straight line of path in the target image
to determine the neighbourhood protein spot
nmlab
14
4. Searching for the protein spot in
the neighbourhood area (2/2)
nmlab
15
5. Selecting the best matched spot
nmlab
16
Experiments and results
1. Identifying a spot along the line of path
2. Identifying a spot of interest in the target
image
3. Matching on 2D gel electrophoresis image
4. Shape comparison
5. Retrieving 3D image
nmlab
17
1. Identifying a spot along the line of
path (1/2)
nmlab
18
1. Identifying a spot along the line of
path (2/2)
threshold
Identifying a spot along the line of path using the
low intensity values.
nmlab
19
2. Identifying a spot of interest in the
target image (1/3)
nmlab
20
2. Identifying a spot of interest in the
target image (2/3)
nmlab
21
2. Identifying a spot of interest in the
target image (3/3)
nmlab
22
3. Matching on 2D gel
electrophoresis image (1/2)
nmlab
23
3. Matching on 2D gel
electrophoresis image (2/2)
nmlab
24
4. Shape comparsion
(a) Source spot
(b) Detected spot
nmlab
25
5. Retrieving 3D image (1/3)
• The following parameters are stored in the
target specific dedicated database:
• Coordinates
• Intensity values
• Positional orientation
• Average shape radius
nmlab
26
5. Retrieving 3D image (2/3)
nmlab
27
5. Retrieving 3D image (3/3)
nmlab
28
Conclusion
• This approach reduces the number of
candidate spot to be identified within the
image.
nmlab
29